search

greearb / ath10k-ct

Stand-alone ath10k driver based on Candela Technologies Linux kernel.
111 stars 40 forks source link

upstream-ath10k pcie DMA discussion #108

Closed MartB closed 4 years ago

MartB commented 4 years ago

Hey there found the following discussion while browsing patches: https://patchwork.kernel.org/patch/11204029/

Looks interesting too me.

greearb commented 4 years ago

That patch did not go upstream yet as far as I can tell, and Felix's original patch claims to fix problems as well. So, maybe this needs more testing or investigation before it is worth applying.

nicklowe commented 4 years ago

The final upstream commit can be found at: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a1769bb68a850508a492e3674ab1e5e479b11254

_"For QCA9984/QCA9888, the DMA_BURST_SIZE register controls the AXI burst size of the RD/WR access to the HOST MEM. 0 - No split , RAW read/write transfer size from MAC is put out on bus as burst length 1 - Split at 256 byte boundary 2,3 - Reserved

With PCIe protocol analyzer, we can see DMA Read crossing 4KB boundary when issue happened. It broke PCIe spec and caused PCIe stuck. So revert the default value from 0 to 1."_

The issue seems to be that a value of 0 means different things depending on the specific chip.

This subsequently has only been changed for the 10_4 target and not others:

TARGET_10_4_DMA_BURST_SIZE 1

greearb commented 4 years ago

I applied this back to 4.16 ath10k-ct drivers (not sure anyone uses anything older). Please pull in latest and test if you can.